README.md

---
license: cc-by-4.0
---
# CoMM: A Coherent Interleaved Image-Text Dataset for Multimodal Understanding and Generation

[![arXiv](https://img.shields.io/badge/arXiv-2406.10462-b31b1b.svg)](https://arxiv.org/abs/2406.10462)
[![Static Badge](https://img.shields.io/badge/Dataset-GoogleDrive-yellow)](https://drive.google.com/file/d/17AWa3wVCk4OZOdBQzDLRXsvRMnaXXP3T/view?usp=sharing)

CoMM is a high-quality dataset designed to improve the coherence, consistency, and alignment of multimodal content. It sources raw data from diverse origins, focusing on instructional content and visual storytelling to establish a strong foundation. 
<img src="data_compare.svg" width="800" alt="data comparison">